DAY 2

library(tidyverse)

── Attaching core tidyverse packages ────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────── tidyverse 2.0.0 ──
✔ dplyr     1.1.2     ✔ readr     2.1.4
✔ forcats   1.0.0     ✔ stringr   1.5.0
✔ ggplot2   3.4.2     ✔ tibble    3.2.1
✔ lubridate 1.9.2     ✔ tidyr     1.3.0
✔ purrr     1.0.1     ── Conflicts ──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────── tidyverse_conflicts() ──
✖ dplyr::filter() masks stats::filter()
✖ dplyr::lag()    masks stats::lag()
ℹ Use the ]8;;http://conflicted.r-lib.org/conflicted package]8;; to force all conflicts to become errors

library(janitor)

Attaching package: ‘janitor’

The following objects are masked from ‘package:stats’:

    chisq.test, fisher.test

steam <- read_csv("raw_data/steam_checkpoint.csv")

Rows: 26564 Columns: 22── Column specification ──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
Delimiter: ","
chr  (8): name, developer, publisher, multiplayer, categories, genres, steamspy_tags, owners
dbl  (8): appid, required_age, achievements, positive_ratings, negative_ratings, average_playtime, median_playtime, price
lgl  (5): free_to_play, virtual_reality_support, windows_support, mac_support, linux_support
date (1): release_date
ℹ Use `spec()` to retrieve the full column specification for this data.
ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.

Where was i?

games_pre_cleaned <- read_csv("raw_data/Cleaned Data 2 GVGS&R.csv")

Rows: 6894 Columns: 15── Column specification ──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
Delimiter: ","
chr  (5): Name, Genre, Publisher, Developer, Rating
dbl (10): Year_of_Release, NA_Sales, EU_Sales, JP_Sales, Other_Sales, Global_Sales, Critic_Score, Critic_Count, User_Score, User_Count
ℹ Use `spec()` to retrieve the full column specification for this data.
ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.

games_pre_cleaned

How dare you do the work for me

games_raw <- read_csv("raw_data/Raw Data GVGS&R.csv")

Rows: 16719 Columns: 16── Column specification ──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
Delimiter: ","
chr (8): Name, Platform, Year_of_Release, Genre, Publisher, User_Score, Developer, Rating
dbl (8): NA_Sales, EU_Sales, JP_Sales, Other_Sales, Global_Sales, Critic_Score, Critic_Count, User_Count
ℹ Use `spec()` to retrieve the full column specification for this data.
ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.

games_raw %>% 
  filter(Name == "Minecraft")

Why does everyone have the same dataset

back to steam i guess

steam

steam %>% 
  distinct(genres)

Already have F2P covered by another column, so we can drop that in genres

steam %>% 
  mutate(genres = str_remove_all(genres, "Free to Play")) %>% 
  filter(free_to_play == TRUE)

Is this worth doing? Duplicated info doesn’t hurt i suppose - leave it for now

single player?

steam <- steam %>% 
  mutate(singleplayer = case_when(
    str_detect(categories, "Single-player") ~ TRUE,
    str_detect(steamspy_tags, "Singleplayer") ~ TRUE,
    TRUE ~ FALSE
  ),.before = multiplayer)

steam %>% 
  filter(free_to_play == TRUE) 

steam

# adding column to simplify ratings 
steam <- steam %>% 
  mutate(total_reviews = positive_ratings +  negative_ratings) %>% 
  arrange(desc(total_reviews)) %>% 
  mutate(percent_positive_reviews = positive_ratings / total_reviews) %>% 
  mutate(general_rating = case_when(
    total_reviews > 1000 & percent_positive_reviews >= 0.95 ~ "Extremely Positive",
    total_reviews > 1000 & percent_positive_reviews >= 0.75 ~ "Positive",
    total_reviews > 1000 & percent_positive_reviews >= 0.55 ~ "Mostly Positive",
    total_reviews > 1000 & percent_positive_reviews >= 0.50 ~ "Mixed" ,
    total_reviews > 1000 & percent_positive_reviews >= 0.35 ~ " Mostly Negative",
    total_reviews > 1000 & percent_positive_reviews >= 0.15 ~ "Negative",
    total_reviews > 1000 & percent_positive_reviews < 0.15 ~ "Extremely Negative",
    positive_ratings == 0 ~ "No Positive Reviews",
    negative_ratings == 0 ~ "No Negative Reviews",
    positive_ratings & negative_ratings == 0 ~ "No Reviews",
    total_reviews < 1000 ~ "Not Enough Reviews",
    TRUE ~ "No Data"
  ),.before = positive_ratings) %>% 
  select(-percent_positive_reviews)

backloggd <- read_csv("clean_data/backloggd_clean.csv")

Rows: 1115 Columns: 15── Column specification ──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
Delimiter: ","
chr  (9): title, team, genre_tag, genre_tag_2, genre_tag_3, genre_tag_4, genre_tag_5, genre_tag_6, genre_tag_7
dbl  (5): years_since_release, rating, number_of_reviews, wishlist, plays
date (1): release_date
ℹ Use `spec()` to retrieve the full column specification for this data.
ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.

backloggd

games_raw <- games_raw %>% 
  clean_names()

# Platform isn't a genre - correcting
games_raw <- games_raw %>% 
  mutate(genre = case_when(
    genre == "Platform" ~ "Platformer",
    TRUE ~ genre
  ))

nintendo_published <- games_raw %>% 
  filter(publisher == "Nintendo")

genre_publisher_sales <- games_raw %>%  
  group_by(publisher, genre) %>% 
  summarise(sum(global_sales)) %>% 
  arrange(desc(`sum(global_sales)`))

`summarise()` has grouped output by 'publisher'. You can override using the `.groups` argument.

write_csv(genre_publisher_sales, "clean_data/genre_by_publisher.csv")

genre_publisher_sales %>% 
  filter(publisher == "From Software")

games_raw %>% 
  distinct(rating)

# clarifying that the rating is for ESRP - the american system
games_raw <- games_raw %>% 
  rename("esrp" = "rating") 

# Adding PEGI equivalent 
games_raw <- games_raw %>% 
  mutate(pegi_equivalent = case_when(
    esrp == "EC" ~ "3",
    esrp == "E" ~ "7",
    esrp == "E10+" ~ "12",
    esrp == "T" ~ "16",
    esrp == "M" ~ "18",
    esrp == "AO" ~ "18",
    esrp == "K-A" ~ "7",
    esrp == "RP" ~ "No Rating",
    TRUE ~ "No Rating Found"
  )) 
  

total_sales_by_genre <- games_raw %>% 
  group_by(genre) %>% 
  summarise(sum(global_sales)) %>% 
  rename("total_sales" = "sum(global_sales)") %>% 
  arrange(desc(total_sales))

total_sales_by_genre <- total_sales_by_genre %>% 
  drop_na()

games_raw 

breakdown_sales_by_genre <- games_raw %>% 
  group_by(genre) %>% 
  summarise(sum(na_sales), sum(eu_sales), sum(jp_sales), sum(other_sales)) %>% 
  rename("total_na_sales"= "sum(na_sales)", "total_eu_sales" = "sum(eu_sales)", "total_jp_sales" = "sum(jp_sales)", "total_other_sales" = "sum(other_sales)")

total_sales_by_genre # this seems to have some rounding differences compared to breakdown, so i'll retire this one 

how does the sales for the top 3 genres look in each region?

breakdown_sales_by_genre <- breakdown_sales_by_genre %>% 
  group_by(genre) %>% 
  mutate(total_sales = sum(total_na_sales + total_eu_sales + total_jp_sales + total_other_sales)) %>% 
  drop_na()

breakdown_sales_by_genre %>% 
  filter(genre %in% c("Action", "Sports", "Shooter")) %>% 
  ggplot(aes(x = total_sales, y = genre))+
  geom_col() +
  geom_col(aes(x = total_na_sales), fill = "green")+
  geom_col(aes(x = total_eu_sales), fill = "red") +
  geom_col(aes(x = total_jp_sales), fill = "blue") 

# this looks shit

What about trends? How did each genres sales adjust every 5 or so years?

games_raw %>% 
  filter(year_of_release == "2016")

games_raw %>% 
  mutate(year_of_release = as.integer(year_of_release)) %>% 
  distinct(year_of_release) %>% 
  arrange(year_of_release)

Warning: There was 1 warning in `mutate()`.
ℹ In argument: `year_of_release = as.integer(year_of_release)`.
Caused by warning:
! NAs introduced by coercion

sales_2019 <- read_csv("raw_data/sales-2019.csv")

Rows: 55792 Columns: 23── Column specification ──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
Delimiter: ","
chr (10): Name, basename, Genre, ESRB_Rating, Platform, Publisher, Developer, Last_Update, url, img_url
dbl (12): Rank, Critic_Score, User_Score, Total_Shipped, Global_Sales, NA_Sales, PAL_Sales, JP_Sales, Other_Sales, Year, status, Vgchartzscore
lgl  (1): VGChartz_Score
ℹ Use `spec()` to retrieve the full column specification for this data.
ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.

sales_2019 %>% 
  distinct(Year) %>% 
  arrange(desc(Year))

sales_2019 %>% 
  filter(Year == "2019") %>% 
  select(Name, Year, Platform)

# data for 2020 is total nonsense. Some information in 2019 is incorrect.

Going off the fact that there is the majority of usable information i am able to get is from 2019 or before, we’re just going to pretend we’re living in an

alternative 2019

As such, I am dropping all data from 2020 in this dataset

sales_2019 <- sales_2019 %>% 
  filter(Year != "2020")

Looking at backloggd again

backloggd

# creating a range_rating column to reduce noise when it comes to plotting results
backlogged <- backloggd %>% 
  mutate(rating_range = case_when(
    rating < 5 & rating >= 4 ~ "4+",
    rating < 4 & rating >= 3 ~ "3 to 4",
    rating < 3 & rating >= 2 ~ "2 to 3",
    rating < 2 & rating >= 1 ~ "1 to 2",
    rating < 1 ~ ">1",
    TRUE ~ "No rating"
  ))

DIFFICULTIES WITH THE DATA

After having little success attempting to get a more up to date data set, i’ve decided to stick with what I already have - Given more time, i would have liked to have run the steam/steam spy script to scrape all the steam data, but unfortunately i couldn’t commit 3 days to it. There doesn’t seem to be much in the way of modern sales data, with most companies not releasing figures unless its to do with a milestone. Genres are quite vague - While it would be nice to have tags like “Open World”, “Rouge-like/lite”, we instead mostly have “Action” “Adventure”, “Indie”, which are not hugely descriptive. This might make recommending a genre to work on harder

---

DAY 3

sales_2019 %>% 
  summary()

      Rank           Name             basename            Genre           ESRB_Rating          Platform          Publisher          Developer         VGChartz_Score  Critic_Score  
 Min.   :    1   Length:54482       Length:54482       Length:54482       Length:54482       Length:54482       Length:54482       Length:54482       Mode:logical   Min.   : 1.00  
 1st Qu.:13638   Class :character   Class :character   Class :character   Class :character   Class :character   Class :character   Class :character   NA's:54482     1st Qu.: 6.40  
 Median :27421   Mode  :character   Mode  :character   Mode  :character   Mode  :character   Mode  :character   Mode  :character   Mode  :character                  Median : 7.50  
 Mean   :27465                                                                                                                                                       Mean   : 7.21  
 3rd Qu.:41202                                                                                                                                                       3rd Qu.: 8.30  
 Max.   :55791                                                                                                                                                       Max.   :10.00  
                                                                                                                                                                     NA's   :47952  
   User_Score    Total_Shipped    Global_Sales      NA_Sales       PAL_Sales        JP_Sales      Other_Sales         Year      Last_Update            url                status 
 Min.   : 2.00   Min.   : 0.03   Min.   : 0.00   Min.   :0.00    Min.   :0.00    Min.   :0.00    Min.   :0.00    Min.   :1970   Length:54482       Length:54482       Min.   :1  
 1st Qu.: 7.80   1st Qu.: 0.20   1st Qu.: 0.03   1st Qu.:0.05    1st Qu.:0.01    1st Qu.:0.02    1st Qu.:0.00    1st Qu.:2000   Class :character   Class :character   1st Qu.:1  
 Median : 8.50   Median : 0.59   Median : 0.12   Median :0.12    Median :0.04    Median :0.05    Median :0.01    Median :2008   Mode  :character   Mode  :character   Median :1  
 Mean   : 8.27   Mean   : 1.89   Mean   : 0.37   Mean   :0.28    Mean   :0.16    Mean   :0.11    Mean   :0.04    Mean   :2006                                         Mean   :1  
 3rd Qu.: 9.20   3rd Qu.: 1.80   3rd Qu.: 0.36   3rd Qu.:0.29    3rd Qu.:0.14    3rd Qu.:0.12    3rd Qu.:0.04    3rd Qu.:2011                                         3rd Qu.:1  
 Max.   :10.00   Max.   :82.86   Max.   :20.32   Max.   :9.76    Max.   :9.85    Max.   :2.69    Max.   :3.12    Max.   :2019                                         Max.   :1  
 NA's   :54154   NA's   :52661   NA's   :35132   NA's   :41528   NA's   :41334   NA's   :47458   NA's   :39005                                                                   
 Vgchartzscore     img_url         
 Min.   :2.60    Length:54482      
 1st Qu.:6.90    Class :character  
 Median :7.90    Mode  :character  
 Mean   :7.49                      
 3rd Qu.:8.50                      
 Max.   :9.60                      
 NA's   :53726                     

All of the VGchartz scores are empty, so it can go. Also, same as above, there’s no such genre as Platform

1970 is nonsense. dropping entire year

ggsave("Visualisations/2019_genres_inc_misc.png")

Saving 7 x 7 in image

 ggsave("Visualisations/2019_genres_excl_misc.png")      

Saving 7 x 7 in image

This is slightly concerning - a lot of these titles could easily have a genre that isn’t the cure-all “Misc” - Include both graphs - one with misc, one excluding it

What about a similar plot for steam games?

steam_years <- steam_years %>%  
  separate(col = genres, sep = ";", into = paste0("Genre", seq_len(max(nchar(.$genres))))) %>%  # this is absolutely horrendous, but it gets me all the genres into separate columns so whatever
  select(-Genre14:-Genre180) %>% 
  pivot_longer(cols = Genre1:Genre13, names_to = "genres") %>% 
  drop_na(value) %>% # ok now we can make the year intervals and group them to find what genres were popular 
  mutate(year_range = case_when(
    release_year <= 2019 & release_year >= 2016 ~ "2016 - 2019",
    release_year <= 2015 & release_year >= 2012 ~ "2012 - 2015",
    release_year <= 2011 & release_year >= 2009 ~ "2009 - 2011",
    release_year <= 2008 & release_year >= 2005 ~ "2005 - 2008",
    release_year <= 2004 & release_year>= 2000 ~ "2000 - 2004",
    release_year < 2001 ~ "Before 2000",
    TRUE ~ "this shouldnt appear"
  ),.after = release_year) 

Warning: Expected 180 pieces. Missing pieces filled with `NA` in 26564 rows [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, ...].

 ggsave("Visualisations/steam_genres_2009_on.png")

Saving 7 x 7 in image

backloggd_genres %>% 
  pivot_longer(cols = genre_tag:genre_tag_7, names_to = "list", values_to = "genre") %>% 
  drop_na(genre)

Error in pivot_longer(., cols = genre_tag:genre_tag_7, names_to = "list",  : 
  object 'backloggd_genres' not found

 ggsave("Visualisations/backloggd_genres.png")

Saving 7 x 7 in image

We have established that Indie, Action and Adventure games seem to be the most popular - But what other elements make them popular?

Note: Action and Adventure are pretty interchangable nothing terms that describe like 90% of titles but oh well

Like would you compare Nier Automata to Donkey Kong Jungle Beat? I wouldn’t but who am i to say

and for comparison

Make a model that tries to guess if a game is positively rated?

{r} # steam_model_base <- steam %>% # separate(col = genres, sep = ";", into = paste0("Genre", seq_len(max(nchar(.$genres))))) %>% # select(-Genre14:-Genre180) #

{r} # steam_model_base %>% # separate(col = categories, sep = ";", into = paste0("categories", seq_len(max(nchar(.$categories))))) %>% # select(-categories21:-categories299) #

{r} # steam %>% # select(categories) %>% # mutate(length_cat = nchar(categories)) %>% # arrange(desc(length_cat)) #

This is idiotic. There are far too many variables. Might mutate a few of these into logicals, like controller support

---

Ok so what do i need to do tomorrow - Continue working on visualisations, maybe make a model that tries to see what factors make games do well on steam?